Security researchers disclosed a vulnerability chain dubbed GrafanaGhost that combined indirect prompt injection with a protocol-relative URL bypass to defeat both client-side protections and Grafana’s AI guardrails. The flaw could have let an attacker silently exfiltrate sensitive dashboard data, such as financial metrics or infrastructure telemetry, without phishing, login credentials, or any visible sign of compromise. Grafana Labs patched the issue across several release branches; researchers followed responsible disclosure and reported no confirmed real-world exploitation against customers.